A prediction-based anti-interference flexible control method for central air conditioning

Abstract

This invention provides a predictive, disturbance-resistant flexible control method for central air conditioning systems, relating to the fields of building energy management and electrical automatic control. It involves constructing a state-space and power nonlinear model of the central air conditioning system; power reference benchmark planning based on nonlinear model predictive control; online parameter identification based on recursive least squares with a forgetting factor; nonlinear decoupling and feedforward conversion of control commands; active estimation of external strong disturbances based on a nonlinear disturbance observer (DOB); online adaptive tuning of PID parameters based on two-dimensional fuzzy logic; and closed-loop control of real-time air conditioning power based on disturbance-resistant Smith prediction. This invention, based on the building's flexible power control requirements, performs nonlinear benchmark planning, time-varying parameter identification, active compensation for strong disturbances, and elimination of pure time delays to achieve closed-loop control of the central air conditioning system, improving the accuracy and response speed of flexible power tracking and assisting in the low-carbon and stable operation of building energy systems.

Description

Technical Field

[0001] This invention relates to the field of building energy management and electrical automatic control, and in particular to a predictive, disturbance-resistant, flexible control method for central air conditioning. Background Technology

[0002] Flexible control of building loads is one of the core supporting technologies for promoting low-carbon operation and ensuring the balance of power grid supply and demand. Centralized air conditioning systems account for a high proportion of total building energy consumption, possessing significant load regulation potential and thermal inertia, making them an important resource for building-grid interaction. Current research on building loads participating in grid regulation mainly focuses on the independent control of central air conditioning systems, while research on their coordinated control with fast-response resources such as electric vehicles and energy storage is relatively limited. Therefore, in scenarios lacking fast-response auxiliary flexible resources and relying solely on central air conditioning as an independent regulating load, improving its refined high-frequency disturbance tracking capability is crucial. The internal electromechanical coupling relationships of central air conditioning systems are complex and have significant time delays. Traditional conventional PID or linear model predictive control is prone to model mismatch, slow response, and system oscillations when facing high-fluctuation power tracking commands and external nonlinear disturbances (such as weather changes and indoor occupant movement), leading to damage to indoor thermal comfort or control failure. Therefore, it is necessary to adopt intelligent nonlinear control algorithms to decouple and compensate for large time delays and strong disturbances within the system. This invention overcomes the large time delay characteristics of centralized air conditioning systems and suppresses nonlinear disturbances by introducing nonlinear model prediction and disturbance-resistant Smith prediction control technology, which meets the urgent needs of new power systems for the accuracy and robustness of building flexible response.

[0003] For power planning in upper-level decision-making, considering the nonlinear time-varying characteristics of the central air conditioning system, a reduced-order thermodynamic and power nonlinear model is first constructed, using predicted solar radiation and internal thermal gain as inputs. A nonlinear model predictive control objective function is established, aiming to minimize grid service tracking error while ensuring the indoor temperature remains within the human comfort range. The optimal power reference trajectory for future periods is obtained by solving within a rolling optimization window. Secondly, for the lower-level execution module, to address the nonlinear disturbance problem caused by drastic environmental changes when the air conditioning system is adjusted independently, a nonlinear disturbance observer is introduced to extract and estimate unmodeled dynamics and external disturbances, feeding them forward to the control input for compensation. Simultaneously, recursive least squares with a forgetting factor is used to update the system's physical parameters in real time. Based on this, fuzzy adaptive Smith predictive PID control is employed, using power tracking error and its rate of change as fuzzy inputs. The PID parameters are adaptively tuned online, and pure time-delay dynamics are separated, thus completing the precise and disturbance-resistant closed-loop control of the central air conditioning system.

[0004] In summary, the flexible control method for central air conditioning based on nonlinear model prediction and disturbance-resistant Smith prediction can effectively complete the nonlinear planning and time delay compensation of the power response of the central air conditioning system under complex disturbances, reduce the dependence on auxiliary fast-response loads, improve the system's disturbance resistance and tracking accuracy, and reduce the risk of control instability caused by model mismatch. This is of great significance for achieving low-carbon and stable operation of buildings. Summary of the Invention

[0005] Based on the above, the purpose of this invention is to provide a prediction-based method for flexible control of central air conditioning systems. Under the premise of obtaining the building's flexible power control requirements, the method performs nonlinear baseline planning, time-varying parameter identification, active compensation for strong disturbances, and elimination of pure time delays to achieve closed-loop control of the central air conditioning system, thereby improving the accuracy and response speed of flexible power tracking and assisting in the low-carbon and stable operation of building energy systems.

[0006] To achieve the above objectives, the present invention provides a prediction-based method for flexible control of central air conditioning systems, the method comprising:

[0007] Step S1: Construct the state-space and power nonlinear model of the central air conditioning system. Establish the building's equivalent heat network, using indoor temperature and wall temperature as state variables, and construct nonlinear coupling equations for chiller unit power consumption, chilled water supply temperature, flow rate, and partial load rate to form a control-oriented nonlinear dynamic model of the central air conditioning system.

[0008] Specifically, it includes:

[0009] Step S11: Extract the thermal inertia characteristics of the building structure, abstract the complex building heat transfer process into three temperature nodes: outdoor environment, building envelope and indoor area, and establish a reduced-order RC thermal equivalent model.

[0010] Step S12: Using indoor temperature and wall temperature as the system's state variables, construct a partial differential thermodynamic state-space equation that includes the input cooling capacity, external ambient temperature, and internal thermal disturbance.

[0011] Step S13: Establish a nonlinear power consumption model for the chiller unit, expressing the real-time power consumption of the central air conditioning system as a nonlinear equation of chilled water supply temperature, flow rate, and chiller unit part load rate (PLR) and coefficient of performance (COP).

[0012] Step S2: Power reference baseline planning based on nonlinear model predictive control. Using the predicted external heat gain and building interior heat gain as disturbance inputs, and under the constraint of room temperature comfort, a rolling window calculation is performed with the objective of minimizing grid service tracking error to plan the optimal power reference trajectory of the central air conditioning system for future periods.

[0013] Specifically, it includes:

[0014] Step S21: Based on weather forecasts and historical load data, extract the solar radiation thermal gain, outdoor ambient temperature, and thermal gain of personnel and equipment inside the building within the prediction time domain as the external known disturbance sequence for the NMPC controller.

[0015] Step S22: Establish the multi-objective optimization cost function of NMPC, including the squared error term for building tracking of grid service instructions (such as absorption or peak shaving), and the transient smoothing term for penalizing drastic changes in control quantity.

[0016] Step S23: Set the nonlinear constraints for NMPC, including the upper and lower limits of the chiller's physical output, flow regulation rate constraints, and indoor thermal comfort constraints. Solve using sequential quadratic programming or the interior point method to output the optimal power target in the future rolling time domain.

[0017] Step S3: Online parameter identification based on recursive least squares with forgetting factor. Real-time power measurements and input feedback from the air conditioning system are collected. The recursive least squares method with forgetting factor is used to identify and dynamically update time-varying physical parameters of the model, such as static gain, time constant, and propagation delay.

[0018] Specifically, it includes:

[0019] Step S31: Collect the operating power measurement value of the chiller unit of the central air conditioning system in real time, and combine it with the control command issued by the controller to form an input-output data pair.

[0020] Step S32: Discretize the thermodynamic process containing large inertia and transform it into a mathematical expression in autoregressive form, and establish the parameter vector to be identified and the data regression vector.

[0021] Step S33: Introduce a forgetting factor to configure the FF-RLS algorithm, perform weight decay on historical outdated data, calculate the gain matrix and update the parameter vector in real time based on the prediction error brought by each sampling, and obtain the latest static gain, time constant and pure lag time parameters.

[0022] Step S4: Nonlinear decoupling and feedforward conversion of control commands. The power reference trajectory planned by the decision layer is sent to the execution module as input. Based on the thermodynamic inverse nonlinear equation of the central air conditioning system, the power tracking target is decoupled and converted into a directly adjustable chilled water supply temperature setpoint and flow reference value.

[0023] Specifically, it includes:

[0024] Step S41: Receive the optimal power target signal issued by the upper-layer NMPC planning.

[0025] Step S42: Based on the chiller unit model, the one-dimensional power command is decoupled and allocated into two setpoints that can be directly executed by the lower-level equipment using a reverse solution mapping or lookup table interpolation algorithm: chilled water supply temperature setpoint and flow rate setpoint.

[0026] Step S5: Active estimation of strong external disturbances based on the nonlinear disturbance observer DOB. To address the issues of insufficient rapid response resources and the impact of weather changes on central air conditioning control, a nonlinear disturbance observer is constructed to extract and estimate unmodeled system dynamics and strong external disturbances in real time, generating feedforward compensation quantities and feeding them forward to the control input.

[0027] Specifically, it includes:

[0028] Step S51: Construct a nominal dynamic model of the controlled object using the latest physical model parameters obtained from parameter identification;

[0029] Step S52: Design a nonlinear disturbance observer, compare the difference between the actual power measurement output of the central air conditioning system and the predicted output of the nominal model, and combine the system state equation to define the unmeasurable environmental changes and internal nonlinear mismatch as equivalent disturbance quantities and perform observation and estimation.

[0030] Step S53: The observed equivalent disturbance is converted into a feedforward compensation signal through a low-pass filter and superimposed on the output of the lower-level controller to achieve active early suppression of external nonlinear interference.

[0031] Step S6: Online adaptive tuning of PID parameters based on two-dimensional fuzzy logic. Using the system's power point tracking error and error rate of change as inputs to the fuzzy logic controller, the proportional, integral, and derivative coefficients of the lower-level PID controller are dynamically corrected in real time according to the designed fuzzy rule base.

[0032] Specifically, it includes:

[0033] Step S61: Calculate the power tracking error e(t) by subtracting the power reference target from the actual measured power, and obtain the error change rate de(t) / dt by differentiating the error. Use both as input variables of the fuzzy controller and perform fuzzification processing to map them into the corresponding fuzzy universe.

[0034] Step S62: Based on experience, formulate a fuzzy rule table containing control logic, focusing on different response stages such as the start-up period, transition period and steady period of the central air conditioning system, and formulate nonlinear adjustment strategies for proportional, integral and derivative corrections respectively.

[0035] Step S63: Use fuzzy inference mechanism to match fuzzy rules, and perform defuzzification calculation through centroid method to output a clear correction gain. Dynamically superimpose this gain onto the initial PID parameters to complete the online adaptive tuning of the PID controller.

[0036] Step S7: Real-time power closed-loop control of air conditioning based on disturbance-resistant Smith prediction. A delay elimination model is constructed using the identified time-varying physical parameters. The pure time delay dynamics caused by the building's large inertia are isolated and eliminated in the predictor feedback loop. By integrating DOB feedforward compensation and fuzzy adaptive PID control, real-time stable power tracking of the central air conditioning equipment layer is achieved.

[0037] Specifically, it includes:

[0038] Step S71: Introduce a Smith predictor structure into the feedback control loop, and use the updated parameters to construct a delay-free equivalent controlled object model to calculate the feedback error in advance, thereby converting the large time-delay system into a time-delay-free system for closed-loop regulation.

[0039] Step S72: The output signal of the fuzzy adaptive PID controller is summed and fused with the feedforward compensation amount generated by the DOB observer to serve as the final physical control command to resist internal and external disturbances.

[0040] Step S73: Send the final physical control command to the chiller unit actuator to drive the unit to operate, and form a closed-loop feedback of real-time energy consumption data to achieve accurate and stable power tracking of the central air conditioning system under disturbance and large delay conditions.

[0041] According to specific embodiments provided by the present invention, the present invention discloses the following technical effects:

[0042] This invention provides a predictive, disturbance-resistant, flexible control method for central air conditioning systems. The method includes: constructing a state-space and power nonlinear model of the central air conditioning system; establishing an equivalent building thermal network, using indoor temperature and wall temperature as state variables, and constructing nonlinear coupling equations for chiller unit power consumption, chilled water supply temperature, flow rate, and partial load rate to form a control-oriented nonlinear dynamic model of the central air conditioning system; planning a power reference baseline based on the nonlinear model predictive control; using predicted external heat gain and building internal heat gain as disturbance inputs, and performing rolling window calculations with the goal of minimizing grid service tracking error while satisfying room temperature comfort constraints to plan the optimal power reference trajectory for the central air conditioning system in future periods; online parameter identification based on recursive least squares with a forgetting factor; real-time acquisition of actual power measurements and input feedback of the air conditioning system, and using recursive least squares with a forgetting factor to identify and dynamically update time-varying physical parameters such as static gain, time constant, and transmission delay of the model in real time; and nonlinear decoupling and feedforward conversion of control commands. The power reference trajectory planned by the decision-making layer is sent to the execution module as input. Based on the thermodynamic inverse nonlinear equation of the central air conditioning system, the power tracking target is decoupled and converted into a directly adjustable chilled water supply temperature setpoint and flow reference value. Active estimation of strong external disturbances is performed based on the nonlinear disturbance observer (DOB). Addressing the lack of fast-response resources and the influence of weather changes on central air conditioning control, a nonlinear disturbance observer is constructed to extract and estimate unmodeled system dynamics and strong external disturbances in real time, generating feedforward compensation and feeding it forward to the control input. Online adaptive tuning of PID parameters is performed based on two-dimensional fuzzy logic. Using the system's power tracking error and error change rate as input to the fuzzy logic controller, the proportional, integral, and derivative coefficients of the lower-level PID controller are dynamically corrected in real time according to the designed fuzzy rule base. Real-time closed-loop power control of the air conditioning system is based on disturbance-resistant Smith prediction. A delay elimination model is constructed using identified time-varying physical parameters. The pure time-delay dynamics caused by the building's large inertia are isolated and eliminated in the predictor feedback loop. By integrating DOB feedforward compensation and fuzzy adaptive PID control, real-time stable power tracking of the central air conditioning equipment layer is achieved. Attached Figure Description

[0043] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0044] Figure 1 is a flowchart of the flexible control method for central air conditioning according to an embodiment of the present invention;

[0045] Figure 2 is a diagram of the anti-interference Smith prediction structure according to an embodiment of the present invention;

[0046] Figure 3 shows some results from an embodiment of the present invention. Detailed Implementation

[0047] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0048] The purpose of this invention is to provide a prediction-based method for flexible control of central air conditioning systems. Under the premise of obtaining the building's flexible power control requirements, the method performs nonlinear baseline planning, time-varying parameter identification, active compensation for strong disturbances, and elimination of pure time delays to achieve closed-loop control of the central air conditioning system, thereby improving the accuracy and response speed of flexible power tracking and assisting in the low-carbon and stable operation of building energy systems.

[0049] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0050] As shown in Figure 1, this invention discloses a prediction-based method for flexible control of central air conditioning systems, the method comprising:

[0051] Step S1: Construct the state-space and power nonlinear model of the central air conditioning system. Establish the building's equivalent heat network, using indoor temperature and wall temperature as state variables, and construct nonlinear coupling equations for chiller unit power consumption, chilled water supply temperature, flow rate, and partial load rate to form a control-oriented nonlinear dynamic model of the central air conditioning system.

[0052] Step S11: Extract the thermal inertia characteristics of the building structure, abstract the complex building heat transfer process into three temperature nodes: outdoor environment, building envelope and indoor area, and establish a reduced-order RC thermal equivalent model.

[0053] Step S12: Using indoor temperature and wall temperature as the system's state variables, construct a partial differential thermodynamic state-space equation that includes the input cooling capacity, external ambient temperature, and internal thermal disturbances. The specific implementation process is as follows:

[0054] To extract the thermal inertia characteristics of building structures, this invention employs an equivalent resistive-capacitive (RC) thermal network to abstract the complex building heat transfer process. Taking indoor temperature T as an example... i and wall temperature T w As state variables, construct the nonlinear partial differential thermodynamic state-space equations: , ;

[0055] Step S13: Establish a nonlinear power consumption model for the chiller unit, expressing the real-time power consumption of the central air conditioning system as a nonlinear equation of chilled water supply temperature, flow rate, and chiller unit part load rate (PLR) and coefficient of performance (COP).

[0056] Step S2: Power reference baseline planning based on nonlinear model predictive control. Using the predicted external heat gain and building interior heat gain as disturbance inputs, and under the constraint of room temperature comfort, a rolling window calculation is performed with the objective of minimizing grid service tracking error to plan the optimal power reference trajectory of the central air conditioning system for future periods.

[0057] Step S21: Based on weather forecasts and historical load data, extract the solar radiation thermal gain, outdoor ambient temperature, and thermal gain of personnel and equipment inside the building within the prediction time domain as the external known disturbance sequence for the NMPC controller.

[0058] Step S22: Establish the multi-objective optimization cost function of NMPC, including the squared error term for building tracking of grid service instructions (such as absorption or peak shaving), and the transient smoothing term for penalizing drastic changes in control quantity.

[0059] Step S23: Set the nonlinear constraints for NMPC, including the upper and lower limits of the chiller's physical output, flow regulation rate constraints, and indoor thermal comfort constraints. Solve using sequential quadratic programming or the interior point method to output the optimal power target in the future rolling time domain. The specific implementation process is as follows:

[0060] The decision-making level needs to ensure that the air conditioning power accurately tracks the adjustment benchmark issued by the power grid while meeting indoor thermal comfort requirements. Figure 3 shows the power optimization results under the system load reduction scenario, within the limited prediction time domain N. p and control time domain N c Within this framework, the objective cost function J for nonlinear model predictive control (NMPC) is established: , In the formula, P HVAC (t+k|t) represents the predicted power of the air conditioning system at time k in the future; P ref ΔU is the power demand command of the power grid; ω1 and ω2 are weighting coefficients used to achieve a balance between power tracking accuracy and control smoothness.

[0061] Step S3: Online parameter identification based on recursive least squares with forgetting factor. Real-time power measurements and input feedback from the air conditioning system are collected. The recursive least squares method with forgetting factor is used to identify and dynamically update time-varying physical parameters of the model, such as static gain, time constant, and propagation delay.

[0062] Step S31: Collect the operating power measurement value of the chiller unit of the central air conditioning system in real time, and combine it with the control command issued by the controller to form an input-output data pair.

[0063] Step S32: Discretize the thermodynamic process containing large inertia and transform it into a mathematical expression in autoregressive form, and establish the parameter vector to be identified and the data regression vector.

[0064] Step S33: Introduce a forgetting factor to configure the FF-RLS algorithm, perform weight decay on historical outdated data, calculate the gain matrix and update the parameter vector in real time based on the prediction error brought by each sampling, and obtain the latest static gain, time constant, and pure lag time parameters. The specific implementation process is as follows:

[0065] Because building thermal parameters and the external environment drift over time, the FF-RLS algorithm is used to update the physical parameters of the first-order inertial plus pure time-delay (FOPDT) model online in real time. The discrete autoregressive equation of the controlled object is: Where θ is the parameter vector to be identified, and φ is the input / output data regression vector. The parameter update rule is as follows:

[0066] (1) Calculate the gain matrix K(t): ;

[0067] (2) Update the parameter vector : ;

[0068] (3) Update the covariance matrix P(t): ;

[0069] Step S4: Nonlinear decoupling and feedforward conversion of control commands. The power reference trajectory planned by the decision layer is sent to the execution module as input. Based on the thermodynamic inverse nonlinear equation of the central air conditioning system, the power tracking target is decoupled and converted into a directly adjustable chilled water supply temperature setpoint and flow reference value.

[0070] Step S41: Receive the optimal power target signal issued by the upper-layer NMPC planning.

[0071] Step S42: Based on the chiller unit model, using a reverse solution mapping or lookup table interpolation algorithm, the one-dimensional power command is decoupled and allocated into two setpoints that can be directly executed by the lower-level equipment: the chilled water supply temperature setpoint and the flow rate setpoint. The specific implementation process is as follows:

[0072] The ideal cooling capacity target Q derived by the decision-makers HVAC Convert to chilled water supply temperature T chw,su With flow rate M chwThe set value. A decoupled equation is established based on the fluid specific heat capacity formula and the chiller unit coefficient of performance (COP), and the formula expression is as follows: , .

[0073] Step S5: Active estimation of strong external disturbances based on the nonlinear disturbance observer DOB. To address the issues of insufficient rapid response resources and the impact of weather changes on central air conditioning control, a nonlinear disturbance observer is constructed to extract and estimate unmodeled system dynamics and strong external disturbances in real time, generating feedforward compensation quantities and feeding them forward to the control input.

[0074] Step S51: Construct a nominal dynamic model of the controlled object using the latest physical model parameters obtained from parameter identification;

[0075] Step S52: Design a nonlinear disturbance observer to compare the difference between the actual power measurement output of the central air conditioning system and the predicted output of the nominal model. Combined with the system state equations, define unmeasurable environmental changes and internal nonlinear mismatches as equivalent disturbances and estimate them through observation. The specific implementation process is as follows:

[0076] To address unmodeled dynamics and strong external disturbances, a nonlinear disturbance observer (DOB) is constructed for feedforward compensation. The actual system output is Y(s), and the nominal model is G. n (s), where the control input is U(s). Equivalent input disturbance The observation equation is defined as: In the formula, Q(s) is a low-pass filter used to suppress high-frequency measurement noise. DOB subtracts the nominal input from the actual equivalent input calculated from the measurement output to obtain an estimate of the disturbance. The system inverts this estimate and directly superimposes it onto the control input to actively cancel the disturbance. .

[0077] Step S53: The observed equivalent disturbance is converted into a feedforward compensation signal through a low-pass filter and superimposed on the output of the lower-level controller to achieve active early suppression of external nonlinear interference.

[0078] Step S6: Online adaptive tuning of PID parameters based on two-dimensional fuzzy logic. Using the system's power point tracking error and error rate of change as inputs to the fuzzy logic controller, the proportional, integral, and derivative coefficients of the lower-level PID controller are dynamically corrected in real time according to the designed fuzzy rule base.

[0079] Step S61: Calculate the power tracking error e(t) by subtracting the power reference target from the actual measured power, and obtain the error change rate de(t) / dt by differentiating the error. Use both as input variables of the fuzzy controller and perform fuzzification processing to map them into the corresponding fuzzy universe.

[0080] Step S62: Based on experience, formulate a fuzzy rule table containing control logic, focusing on different response stages such as the start-up period, transition period and steady period of the central air conditioning system, and formulate nonlinear adjustment strategies for proportional, integral and derivative corrections respectively.

[0081] Step S63: Fuzzy rules are matched using a fuzzy inference mechanism, and defuzzification calculations are performed using the centroid method to output a clear correction gain. This gain is then dynamically superimposed onto the initial PID parameters to complete the online adaptive tuning of the PID controller. The specific implementation process is as follows:

[0082] The fuzzy inference engine outputs the PID parameter increment ΔK based on the membership degrees of the error E and the error change rate EC. p ,ΔK i ,ΔK d The fuzzy set. To obtain precise physical control quantities, the centroid method is used for defuzzification: ; where μ j c is the matching degree of the j-th activated fuzzy rule; j Let be the center value of the consequent of the j-th rule in the universe of discourse. Real-time PID parameters are generated by superimposing nominal values ​​and increments, thereby dynamically adjusting the controller stiffness according to different response stages of the large time-delay system.

[0083] Step S7: Real-time power closed-loop control of air conditioning based on disturbance-resistant Smith prediction. A delay elimination model is constructed using the identified time-varying physical parameters. The pure time delay dynamics caused by the building's large inertia are isolated and eliminated in the predictor feedback loop. By integrating DOB feedforward compensation and fuzzy adaptive PID control, real-time stable power tracking of the central air conditioning equipment layer is achieved.

[0084] Step S71: Introduce a Smith predictor structure into the feedback control loop. Use the updated parameters to construct a time-delay-free equivalent controlled object model to calculate the feedback error in advance, thereby converting the large time-delay system into a time-delay-free system for closed-loop regulation, as shown in Figure 2. The specific implementation process is as follows:

[0085] Traditional closed-loop error calculations for central air conditioning systems, which suffer from pure time delay, often result in severe overshoot. This invention utilizes the Smith predictor principle to introduce a time-delay-free model into the feedback loop. and time delay model The difference is used to correct the error. The corrected prediction error is expressed in the time domain as: .

[0086] This compensation mechanism feeds back the power change that has not yet been reflected in the actual output due to time delay to the adaptive PID controller in advance, eliminating the pure time delay term in the closed-loop characteristic equation and improving the stability margin of the system under high-frequency control.

[0087] Step S72: The output signal of the fuzzy adaptive PID controller is summed and fused with the feedforward compensation amount generated by the DOB observer to serve as the final physical control command to resist internal and external disturbances.

[0088] Step S73: Send the final physical control command to the chiller unit actuator to drive the unit to operate, and form a closed-loop feedback of real-time energy consumption data to achieve accurate and stable power tracking of the central air conditioning system under disturbance and large delay conditions. Figure 3 shows the power tracking results in the business scenario, indicating that the method shown can achieve effective tracking of the target power.

[0089] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on the differences from other embodiments. The same or similar parts between the various embodiments can be referred to each other.

[0090] This document uses specific examples to illustrate the principles and implementation methods of the present invention. The descriptions of the above embodiments are only for the purpose of helping to understand the method and core ideas of the present invention. Furthermore, those skilled in the art will recognize that, based on the ideas of the present invention, there will be changes in the specific implementation methods and application scope. Therefore, the content of this specification should not be construed as a limitation of the present invention.

Claims

1. A prediction-based method for flexible control of central air conditioning systems, characterized in that, Including the following steps: Step S1: Construct the state space and power nonlinear model of the central air conditioning system; establish the building equivalent heat network, with indoor temperature and wall temperature as state variables, and construct the nonlinear coupling equations of chiller unit power consumption, chilled water supply temperature, flow rate and partial load rate to form a control-oriented nonlinear dynamic model of central air conditioning. Step S2: Power reference baseline planning based on nonlinear model predictive control; using the predicted external heat gain and building internal heat gain as disturbance inputs, and under the constraint of room temperature comfort, performing rolling window calculation with the goal of minimizing grid service tracking error, to plan the optimal power reference trajectory of the central air conditioning system for future periods; Step S3: Online parameter identification based on recursive least squares method with forgetting factor; The system collects real power measurements and input feedback from the air conditioning system in real time, and uses a recursive least squares method with a forgetting factor to identify and dynamically update time-varying physical parameters such as static gain, time constant, and transmission delay of the model in real time. Step S4: Nonlinear decoupling and feedforward conversion of control commands; The power reference trajectory planned by the decision-making level is sent to the execution module as input. Based on the thermodynamic inverse nonlinear equation of the central air conditioning system, the power tracking target is decoupled and converted into a directly adjustable chilled water supply temperature setpoint and flow reference value. Step S5: Active estimation of strong external disturbances based on nonlinear disturbance observer DOB; In response to the lack of fast-response resources and the impact of weather changes on central air conditioning control, a nonlinear disturbance observer is constructed to extract and estimate unmodeled dynamics and strong external disturbances in the system in real time, generate feedforward compensation quantities and feed them forward to the control input. Step S6: Online adaptive tuning of PID parameters based on two-dimensional fuzzy logic; using the system's power tracking error and error change rate as the input of the fuzzy logic controller, and dynamically correcting the output of the proportional, integral, and derivative coefficients of the lower-level PID controller in real time according to the designed fuzzy rule base; Step S7: Real-time power closed-loop control of air conditioning based on disturbance-resistant Smith prediction; using the identified time-varying physical parameters to construct a delay elimination model, the pure time delay dynamics caused by the large inertia of the building are isolated and eliminated in the predictor feedback loop, and DOB feedforward compensation and fuzzy adaptive PID control are integrated to achieve real-time stable power tracking of the central air conditioning equipment layer.

2. The predictive-based flexible control method for central air conditioning systems according to claim 1, characterized in that, Step S1 specifically includes: Step S11: Extract the thermal inertia characteristics of the building structure, abstract the complex building heat transfer process into three temperature nodes: outdoor environment, building envelope and indoor area, and establish a reduced-order RC thermal equivalent model. Step S12: Using indoor temperature and wall temperature as the system's state variables, construct a partial differential thermodynamic state-space equation that includes the input cooling capacity, external ambient temperature, and internal thermal disturbance. Step S13: Establish a nonlinear power consumption model for the chiller unit, expressing the real-time power consumption of the central air conditioning system as a nonlinear equation of chilled water supply temperature, flow rate, chiller unit partial load rate (PLR), and coefficient of performance (COP).

3. The predictive-based flexible control method for central air conditioning systems according to claim 1, characterized in that, Step S2 specifically includes: Step S21: Based on weather forecasts and historical load data, extract the solar radiation heat gain, outdoor ambient temperature, and building personnel and equipment heat gain in the prediction time domain as the external known disturbance sequence for the NMPC controller. Step S22: Establish the multi-objective optimization cost function of NMPC, including the squared error term for building tracking power grid service instructions, and the transient smoothing term for penalizing drastic changes in control quantity; Step S23: Set the nonlinear constraints of NMPC, including the physical output upper and lower limits of the chiller unit, the flow regulation rate constraint, and the indoor thermal comfort constraint; solve the problem by sequential quadratic programming or interior point method, and output the optimal power target in the future rolling time domain.

4. The predictive-based flexible control method for central air conditioning systems according to claim 1, characterized in that, Step S3 specifically includes: Step S31: Collect the operating power measurement value of the chiller unit of the central air conditioning system in real time, and form an input-output data pair with the control command issued by the controller; Step S32: Discretize the thermodynamic process containing large inertia and transform it into a mathematical expression in autoregressive form, and establish the parameter vector to be identified and the data regression vector; Step S33: Introduce a forgetting factor to configure the FF-RLS algorithm, perform weight decay on historical outdated data, calculate the gain matrix and update the parameter vector in real time based on the prediction error brought by each sampling, and obtain the latest static gain, time constant and pure lag time parameters.

5. The predictive-based flexible control method for central air conditioning systems according to claim 1, characterized in that, Step S4 specifically includes: Step S41: Receive the optimal power target signal issued by the upper-layer NMPC planning; Step S42: Based on the chiller unit model, the one-dimensional power command is decoupled and allocated into two setpoints that can be directly executed by the lower-level equipment using a reverse solution mapping or lookup table interpolation algorithm: chilled water supply temperature setpoint and flow rate setpoint.

6. The predictive-based flexible control method for central air conditioning systems according to claim 1, characterized in that, Step S5 specifically includes: Step S51: Construct a nominal dynamic model of the controlled object using the latest physical model parameters obtained from parameter identification; Step S52: Design a nonlinear disturbance observer, compare the difference between the actual power measurement output of the central air conditioning system and the predicted output of the nominal model, and combine the system state equation to define the unmeasurable environmental changes and internal nonlinear mismatch as equivalent disturbance quantities and perform observation and estimation. Step S53: The observed equivalent disturbance is converted into a feedforward compensation signal through a low-pass filter and superimposed on the output of the lower-level controller to achieve active early suppression of external nonlinear interference.

7. The predictive-based flexible control method for central air conditioning systems according to claim 1, characterized in that, Step S6 specifically includes: Step S61: Calculate the power tracking error e(t) by subtracting the power reference target from the actual measured power, and obtain the error change rate de(t) / dt by differentiating the error. Use both as input variables of the fuzzy controller and perform fuzzification processing to map them into the corresponding fuzzy universe. Step S62: Based on experience, formulate a fuzzy rule table containing control logic, focusing on different response stages such as the start-up period, transition period and steady period of the central air conditioning system, and formulate nonlinear adjustment strategies for proportional, integral and derivative corrections respectively; Step S63: Use fuzzy inference mechanism to match fuzzy rules, and perform defuzzification calculation through centroid method to output a clear correction gain. Dynamically superimpose this gain onto the initial PID parameters to complete the online adaptive tuning of the PID controller.

8. The predictive-based flexible control method for central air conditioning systems according to claim 1, characterized in that, Step S7 specifically includes: Step S71: Introduce a Smith predictor structure into the feedback control loop, and use the updated parameters to construct a delay-free equivalent controlled object model to calculate the feedback error in advance, thereby converting the large time-delay system into a time-delay-free system for closed-loop regulation. Step S72: The output signal of the fuzzy adaptive PID controller is summed and fused with the feedforward compensation amount generated by the DOB observer to serve as the final physical control command to resist internal and external disturbances. Step S73: Send the final physical control command to the chiller unit actuator to drive the unit to operate, and form a closed-loop feedback of real-time energy consumption data to achieve accurate and stable power tracking of the central air conditioning system under disturbance and large delay conditions.